Dysregulated ΔNp63α Inhibits Expression of Ink4a/arf, Blocks Senescence, and Promotes Malignant Conversion of Keratinocytes

p63 is critical for squamous epithelial development, and elevated levels of the ΔNp63α isoform are seen in squamous cell cancers of various organ sites. However, significant controversy exists regarding the role of p63 isoforms as oncoproteins or tumor suppressors. Here, lentiviruses were developed to drive long-term overexpression of ΔNp63α in primary keratinocytes. Elevated levels of ΔNp63α in vitro promote long-term survival and block both replicative and oncogene-induced senescence in primary keratinocytes, as evidenced by the expression of SA-β-gal and the presence of nuclear foci of heterochromatin protein 1γ. The contribution of ΔNp63α to cancer development was assessed using an in vivo grafting model of experimental skin tumorigenesis that allows distinction between benign and malignant tumors. Grafted lenti-ΔNp63α keratinocytes do not form tumors, whereas lenti-GFP/v-rasHa keratinocytes develop well-differentiated papillomas. Lenti-ΔNp63α/v-rasHa keratinocytes form undifferentiated carcinomas. The average volume of lenti-ΔNp63α/v-rasHa tumors was significantly higher than those in the lenti-GFP/v-rasHa group, consistent with increased BrdU incorporation detected by immunohistochemistry. The block in oncogene-induced senescence corresponds to sustained levels of E2F1 and phosphorylated AKT, and is associated with loss of induction of p16ink4a/p19arf. The relevance of p16ink4a/p19arf loss was demonstrated in grafting studies of p19arf-null keratinocytes, which develop malignant carcinomas in the presence of v-rasHa similar to those arising in wildtype keratinocytes that express lenti-ΔNp63α and v-rasHa. Our findings establish that ΔNp63α has oncogenic activity and its overexpression in human squamous cell carcinomas contributes to the malignant phenotype, and implicate its ability to regulate p16ink4a/p19arf in the process.

Introduction p63 is a p53 homologue. The p63 gene contains the three functional domains homologous to those of p53, which mediate transactivation (TA), DNA binding (DBD) and oligomerization (OD) [1]. However, in contrast to p53's well established role as a tumor suppressor [2], p63 has been primarily considered a critical developmental regulator of epithelium. It is well understood that temporal regulation of individual p63 isoforms is required for both normal development and maintenance of mature epidermis. This is evidenced by studies in p63 null mice, which are born with severe abnormalities, including the lack of epidermis and many ectodermal derivatives, truncated limbs and craniofacial malformations [3,4] and further supported by studies of postnatal keratinocytes in which p63 isoforms have been manipulated [5,6].
Despite similarities in their structures, p63 is also distinct from p53 in its role in tumorigenesis. While p53 is one of the most commonly mutated genes recognized to date in human malignancies, p63 is rarely mutated in human cancers [7], though p63 gene amplification and/or overexpression has been reported in human squamous cell carcinomas (SCC) of the head and neck, lung, cervix and skin [1,[8][9][10][11][12].
p63 is further distinct from p53 in its role in cell senescence. It is now well appreciated that senescence represents a potent anticancer mechanism to prevent tumor progression from premalignant to malignant lesions [13,14]. In contrast to p53's established role in promoting this tumor-suppressive machinery, it has been shown that p63 deficiency leads to the activation of cell senescence and accelerated aging in mice [15].
Significant controversy exists regarding the role of p63 as an oncogene or as a tumor suppressor gene [7]. In a study by Flores et al. [16], p63 +/2 mice were found to have increased susceptibility to spontaneous tumorigenesis. A complex tumor phenotype was observed in the mutant mice, which included squamous cell carcinomas, histiocytic sarcomas and adenomas. Mice heterozygous for null mutations in both p53 and p63 developed higher tumor burdens and had higher rates of metastases compared to p63 +/+ /p53 +/2 mice. These findings indicate that loss of p63 may cooperate in tumor formation with p53 loss-of-function. In contrast, an independent study by Keyes et al. [17] reported that p63 +/2 mice were less prone to either spontaneous or chemically induced tumors. The neoplasms that did develop in p63+/2 mice included lymphomas, sarcomas and carcinomas. In the latter study, mice heterozygous for null mutations in both p63 and p53 were found to have fewer tumors than p63 +/+ /p53 +/2 mice. These findings suggest that loss of p63 may prevent tumor formation mediated by p53 loss-of-function.
The complexity of p63 contributes to the confusion surrounding the role of p63 in tumorigenesis [7]. p63 protein may refer to multiple variants arising from alternate promoter usage and/or alternative splicing. The p63 gene is transcribed into two subclasses, TA and DN, which differ at the amino-terminus [1]. Additionally, alternative splicing gives rise to COOH-terminal variants p63a, -b and -c within both TA-and DN-subclasses. TAp63 isoforms contain a p53-like N-terminal transactivation (TA) domain and are capable of transactivating known p53responsive genes. DNp63 isoforms are transcribed from an alternate promoter and lack this transactivation domain, while still retaining transactivation activity [1,[18][19][20]. DN isoforms have also been shown to be capable of acting in a dominant-negative manner to block transactivation mediated by TAp63 isoforms as well as by p53 [1].
Accumulating evidence implies that the TA and DN isoforms have distinct or even opposing functions in neoplasia. Although it has been suggested that the tumor suppressor phenotype of p63 might come from TAp63 but not DNp63 isoforms [18], significant controversy still exists regarding the role of individual p63 isoforms in tumorigenesis. Decreased TAp63 levels have been linked to poor clinical outcomes in buccal and laryngeal squamous cell carcinomas [21,22]. TAp63 functions as a robust mediator of cell senescence and inhibits tumorigenesis in vivo [23]. TAp63 isoforms have been found to promote cell apoptosis through death receptors and activating proapoptotic Bcl-2 family members [24]. A role of TAp63 in DNA damage-induced cell cycle arrest and cell death has also been demonstrated [25]. Conversely, Koster et al. reported, using an inducible transgenic mouse model, that embryonic induction of TAp63a causes keratinocyte hyperproliferation, and inhibits terminal differentiation [26], and that post-natal induction of TAp63a accelerates tumor development and progression [27].
DNp63 has been reported to be over-expressed in several different human cancers [8,12]. DNp63a is the predominant isoform in basal keratinocytes, and its expression correlates with proliferation in human cancers [28]. DNp63a has been shown to promote survival in squamous epithelial malignancy by repressing a p73-dependent proapoptotic transcriptional program [29]. However, it has also been reported that DNp63a promotes basal keratinocyte withdrawal from the cell cycle and commitment to terminal differentiation or apoptosis [30,31].
We have previously reported that transient overexpression of DNp63a causes enhanced cell proliferation and inhibition of morphological and biochemical differentiation in primary mouse keratinocytes [20,32]. The long-term biological effects of DNp63a and the consequences of it's overexpression, as seen in human squamous cell carcinomas, remain unknown. In this study we apply a well-established model of squamous cancer to elucidate the role of elevated levels of p63 in distinct stages of tumorigenesis. We report that sustained elevated levels of DNp63a in normal cells block the p16 ink4a /p19 arf pathways and promote keratinocyte survival. However, elevated levels of DNp63a alone are insufficient to confer a tumor phenotype in vivo. In addition to its effect on replicative senescence, elevated DNp63a blocks oncogene-induced senescence and associated induction of p16 ink4a /p19 arf , and cooperates with v-ras Ha to enhance malignant conversion in vivo. These findings indicate an oncogenic role for the elevated levels of DNp63a that have been observed in human squamous cell cancers.

Lentivirus infection drives long term and stable gene expression in primary keratinocytes
To mimic the sustained high level expression of DNp63 that has been observed in human SCC, lentiviruses encoding green fluorescent protein (lenti-GFP) or DNp63a (lenti-DNp63a) were developed and their ability to drive long-term gene expression was assessed in cultures of primary epidermal keratinocytes. Under optimized conditions, the infection efficiency as assessed with lenti-GFP was 92.05% in primary keratinocytes five days following lentiviral infection ( Figure 1A). Lenti-DNp63a expression was monitored over time by western blotting ( Figure 1B). At five days following infection, lenti-DNp63a levels resembled those at the peak of adenoviral-driven gene expression (days 2-3, left panel), and remained stable for at least 14 days (right panel). Consistent with the FACS analysis shown for lenti-GFP in figure 1A, lenti-DNp63a expression observed by immunohistochemistry was distributed uniformly across the cell population ( Figure 1C).

Long-term DNp63a overexpression promotes survival and blocks replicative senescence of primary keratinocytes
The effect of elevated levels of DNp63a on long-term in vitro survival of primary keratinocyte cultures was assessed by microscopic evaluation over 14 days following introduction of lenti-GFP or lenti-DNp63a. As shown in Figure 2A, lenti-DNp63a cultures remain confluent and appeared healthier at 14 days compared to lenti-GFP control cultures. Proliferation status was assessed by BrdU uptake followed by FACS analysis ( Figure 2B). The S-phase population was initially similar across the two cell populations, as seen 5 days following introduction of the lentiviruses (18.18% vs. 17.58% in lenti-DNp63a-vs. lenti-GFPexpressing keratinocytes, respectively). While the S-phase population declined in both cell populations over 2 weeks in culture, it was maintained at higher levels in lenti-DNp63a cultures vs. the parallel lenti-GFP control cultures (11.12% vs. 4.12% BrdUpositive cells, respectively, at 10 days; 5.12% vs. 1.59% BrdUpositive cells, respectively, at day 14).
The continually decreasing BrdU uptake levels in both lenti-DNp63a and control cultures indicate that other mechanisms, in addition to cell proliferation, are involved in the prolonged maintenance of DNp63a overexpressing cells. Therefore, we tested the notion that overexpression of DNp63a facilitates the capacity of the cell to overcome replicative senescence, as originally described by Hayflick in cultured cells [33]. A previous study by Keyes et al., showed that inactivation of all p63 isoforms in mice lead to cellular senescence and accelerated aging [15]. However, the role of individual p63 isoforms in cellular senescence has remained unclear. Here, keratinocytes were infected with lenti-GFP or lenti-DNp63a, and the cell senescence status at day 14 was determined by the nuclear presence of prominent HP1c-positive senescence-associated heterochromatin foci [34]. As shown in Figure 2C, distinct HP-1c nuclear foci were observed in control keratinocytes that express lenti-GFP, but not in keratinocytes that express lenti-DNp63a. Double staining with both HP-1c and p63 antibodies demonstrated that the formation of HP-1c foci is associated with lower p63 expression ( Figure 2C, arrows).
Senescence in most cells is regulated through some combination of activities within the RB and p53 pathways, but frequently operates in a context-dependent and complex manner [13]. We examined the status of the RB and p53 pathways in keratinocytes that differ in their senescence status due to expression of either lenti-GFP or lenti-DNp63a. The induction of p53, p21, p16 ink4a , p19 arf and E2F1 were examined by western blot. No induction of p53 or p21 was observed during the timepoints studied (data not shown). In contrast, up-regulation of p19 arf followed by p16 ink4a was observed beginning by day 3 in control lenti-GFP keratinocytes cultures ( Figure 3A). This expression of p16 ink4a and p19 arf increased over time in lenti-GFP cultures, but was attenuated and delayed in lenti-DNp63a-expressing keratinocytes ( Figure 3A). As expected, the upregulation of p16 ink4a and p19 arf correlates with decreased levels of E2F1. The down-regulation of E2F1 was delayed and attenuated in cells over-expressing DNp63a  ( Figure 3A). We further used RT-PCR to confirm that the regulation of p16 ink4a and p19 arf by DNp63a occurs at the transcriptional level. RT-PCR analysis on p16 ink4a and p19 arf transcription was carried out on primary keratinocytes at day 3, 5, 7 and 10 post-lenti-GFP or lenti-DNp63a infection. As shown in figure 3B, p16 ink4a and p19 arf mRNA began to increase in lenti-GFP cells at day 5 post-infection. This induction was delayed in lenti-DNp63a overexpressing cells. These findings are consistent with a previous report, using cells null for all p63 isoforms, demonstrating that p63 directly represses p16 Ink4a and p19 Arf expression [35]. To further challenge the relationship between DNp63a and p16 ink4a /p19 arf , a transient induction of DNp63a was achieved using adenoviral-mediated gene transduction and used to assess its impact on temporal expression of p16 ink4a /p19 arf . Overexpression of DNp63a inhibits the up-regulation of p16 ink4a and p19 arf associated with cell senescence. A, Whole cell protein was collected from primary keratinocytes expressing lenti-GFP or lenti-DNp63a at timepoints indicated following lentiviral gene transduction. Expression levels of p16 ink4a , p19 arf , E2F1 and p63 were detected by western blot. Equal protein loading was confirmed by immunoblotting for bactin. B, Primary keratinocytes were infected with lentivirus encoding DNp63a or GFP at day 3 after plating. Total RNA was harvested at day 3, 5, 7 and 10 post-infection and reverse transcribed. Expression of p16 ink4a and p19 arf was determined by PCR amplification. PCR amplification of GAPDH was used as a loading control. C, Primary keratinocytes were cultured for 7 days after plating and then infected with adenovirus encoding DNp63a or LacZ. Whole cell lysates were collected at day 2, 3, 6 and 8 post-adenovirus infection (equivalent to day 9, 10, 13 and 15 post-plating). The expression levels of p16 ink4a , p19 arf and p63 were detected by western blot. Equivalent protein loading was confirmed by immunoblotting for actin. doi:10.1371/journal.pone.0021877.g003 Primary keratinocytes were infected with adenoviral constructs encoding Lac-Z or DNp63a at day 7 after plating, when p16 ink4a and p19 arf levels normally begin to increase. The expression levels of p16 ink4a , p19 arf and p63 were evaluated at days 2, 3, 6 and 8 after adenoviral infection (equivalent to days 9, 10, 13 and 15 after plating, and to the time point of day 6, 7, 10 and 12 following introduction of lentivirus in Figure 3A). As shown in Figure 3C, adenoviral-driven gene expression of DNp63a peaked at day 2 post-infection and then rapidly declined. In control Ad-lacZ cultures, both p16 ink4a and p19 arf levels were seen to increase 2 to 3 days following adenovirus infection, in association with the onset of senescence typically seen in these cultures 10 days after plating. This upregulation was abrogated in parallel cultures in which Ad-DNp63a had been introduced. As Ad-DNp63a levels declined, levels of both p16 ink4a and p19 arf were once again up-regulated.
Elevated DNp63a levels enhance malignant conversion in v-ras Ha -expressing keratinocytes The role of DNp63a in promoting cell proliferation and blocking the p16 ink4a /p19 arf pathways of cell senescence, as shown in Figures 2 and 3, indicate that DNp63a might function as an oncogene in carcinogenesis. To determine the contribution of DNp63a to cancer pathogenesis, we applied a well established in vivo grafting model of experimental skin cancer that allows distinctions between benign and malignant tumor phenotypes. We introduced lenti-GFP or lenti-DNp63a into primary murine keratinocytes, either alone or in combination with retrovirus encoding a v-ras Ha oncogene, and cultures were grafted onto the dorsal surface of nude mice in combination with cultured dermal fibroblasts as previously described [36]. No tumors were observed in grafts of lenti-GFP or lenti-DNp63a keratinocytes following grafting ( Figure 4A and Table 1). Graft sites on 4 of the 15 animals that received lenti-GFP/v-ras Ha -expressing keratinocytes developed benign, well-differentiated papillomas with an average tumor volume of 25 mm 3 . In contrast, when v-ras Ha was expressed in combination with lenti-DNp63a, all grafts (20 mice) gave rise to undifferentiated carcinomas with an average tumor volume of 831.9 mm 3 (figure 4A, B and Table 1). The increased tumor size corresponded to increased BrdU incorporation, as detected by immunohistochemistry ( Figure 4B and C), reflecting a higher proliferation status in these tumors.
Long term DNp63a overexpression blocks oncogeneinduced senescence and supports long term survival of v-ras Ha -expressing primary keratinocytes It has been well documented that normal keratinocytes undergo a transient hyperproliferation response followed by senescence after oncogenic v-ras Ha activation [37]. Oncogene-induced senescence is now well appreciated to be a crucial barrier to malignant conversion [13]. The above results (Figure 4) demonstrate that elevated DNp63a levels enhance malignant conversion in v-ras Ha -expressing keratinocytes and suggest that DNp63a may play a role in blocking oncogene-induced senescence as well as replicative senescence. Primary keratinocytes were transduced with retrovirus encoding oncogenic v-ras Ha followed by lentivirus encoding either lenti-GFP ( Figure 5A, left) or lenti-DNp63a ( Figure 5A, right). Oncogene-induced senescence was demonstrated at day 14 with the presence of nuclear foci of HP-1c ( Figure 5A, upper panels) or SA-b-Gal staining ( Figure 5A, lower panels). While both of these well-defined markers of senescence were observed in lenti-GFP/v-ras Ha cultures, they were diminished in lenti-DNp63a/v-ras Ha keratinocytes. p53 has been shown to direct cell fate between quiescence vs. senescence, and this has been linked to the status of the mTOR pathway [38,39]. We evaluated levels of phospho-S6 as a marker of mTOR activity in v-ras Ha -expressing primary keratinocytes that had been transduced with either lenti-GFP or lenti-DNp63a. No change was observed in phospho-S6 levels due to the DNp63a status at days 5, 7 and 14 post-lentiviral infection, the timeframe during which control cultures are transitioning through senescence. These findings indicate that suppression of senescence in this system is not mediated via inhibition of mTOR (data not shown).
Similar to results seen with keratinocytes undergoing replicative senescence ( Figure 3A), western blotting revealed up-regulation of both p16 ink4a and p19 arf in the presence of v-ras Ha (lenti-GFP/vras Ha ) that was attenuated and delayed in the additional presence of elevated DNp63a ( Figure 5B). Furthermore, consistent with the increased proliferation status observed in tumors generated from lenti-DNp63a/v-ras Ha keratinocytes ( Figure 4B and C), the levels of E2F1 and phosphorylated AKT (serine 473) remain stable in lenti-DNp63a/v-ras Ha keratinocytes, but start to decrease at day 7 in lenti-GFP/v-ras Ha keratinocytes ( Figure 5B).
We have shown that sustained dysregulation of DNp63a in keratinocytes supports cell proliferation and facilitates an escape from oncogene-induced senescence ( Figure 5A and B). We tested whether co-expression of DNp63a and v-ras Ha would cooperate to immortalize cells and allow multiple passaging of primary mouse keratinocytes. Primary cultures of v-ras Ha -expressing keratinocytes were transduced with lentiviruses encoding lenti-GFP or lenti-DNp63a. The cells were trypsinized 14 days following lentivirus introduction and reseeded. As shown in Figure 6A, lenti-GFP/vras Ha keratinocytes display a senescence-like morphology 3 days after replating that is not observed in lenti-DNp63a/v-ras Ha keratinocytes. Furthermore, lenti-DNp63a/v-ras Ha keratinocytes re-attached and expanded for at least two passages, whereas lenti-GFP/v-ras Ha keratinocytes stopped proliferation after the initial reseeding ( Figure 6B).

Loss of p19 arf cooperates with v-ras Ha in malignant conversion, similar to lenti-DNp63a-expressing keratinocytes
To determine whether DNp63a-mediated downregulation of p16 ink4a or p19 arf could enhance malignant conversion, p19 arf -null primary keratinocytes were transduced with v-ras Ha -encoding retrovirus, followed by lenti-GFP or lenti-DNp63a and grafted onto the dorsal surface of nude mice as previously described [36]. Wildtype primary keratinocytes expressing lenti-GFP or lenti-DNp63a in combination with v-ras Ha were used as negative and positive controls. Out of 10 animals grafted with p19 arf -null keratinocytes expressing lenti-GFP and v-ras Ha , 9 animals gave rise to undifferentiated carcinomas, which appeared similar to the tumor phenotype derived from keratinocytes overexpressing DNp63a. The remaining animal developed a well differentiated papilloma ( Figure 7A and B, Table 2). A significantly larger tumor volume was observed in tumors derived from lenti-DNp63a/vras Ha -expressing p19 arf -null keratinocytes when compared to tumors derived from p19 arf -null cells that express v-ras Ha alone ( Figure 7B).

Discussion
Expression of DNp63a is associated with proliferation of both normal and neoplastic epithelium, with elevated levels observed in human squamous cancer tissues [1,[8][9][10][11][12]. Here, we provide a mechanistic link between the elevated levels of DNp63a and development of squamous cell cancers. The grafting model applied in this study is relevant to the normal tissue distribution of p63, and allows the discrimination of genetic alterations that contribute to distinct stages of cancer development. We demonstrate that DNp63a overexpression alone does not confer a neoplastic phenotype upon normal primary keratinocytes under the condi-tions used, but enhances malignant conversion of benign tumors arising from primary murine keratinocytes that express oncogenic ras. This phenotypic change appears to be due, at least in part, to the ability of DNp63a to enhance cell survival through suppression of p16 ink4a /p19 arf , key mediators of cell senescence. Cell senescence was initially described as a form of irreversible growth arrest in cultured human cells [33]. It is now understood that it can be prematurely induced by multiple stimuli including DNA damage, oxidative stress, and excessive mitogenic stimuli [40], [41], and represents a potent anticancer mechanism to prevent malignant conversion [14]. The complexities of the signaling pathways mediating the senescence response, and the dependence of this response on p53, are underscored by the contrasting results observed in vitro and in vivo in keratinocytes undergoing senescence induced by p63 ablation [15].
The inhibition of cell senescence by DNp63a is linked to its ability to repress the expression of p16 ink4a and p19 arf ( Figure 5). While p16 ink4a and p19 arf play well-defined roles in controlling cell cycle and senescence, their regulation remains poorly understood. Our finding that DNp63a negatively regulates p16 ink4a and p19 arf in primary mouse keratinocytes is consistent with recent studies linking p63 to cellular senescence and organismal aging, in which keratinocytes of p63-deficient mice display increased levels of p16 ink4a /p19 arf [15]. The ability of p63 to bind putative p53/p63 consensus sites on the p16 ink4a /p19 arf promoters, as demonstrated by ChIP analysis using cells null for all isoforms of p63, raises the possibility that p63 acts as a direct repressor of the p16 ink4a /p19 arf locus. A function for p16 ink4a downstream of p63 is also supported by the observation that germline disruption of both p16 ink4a and p19 arf was found to significantly alleviate the phenotypic consequences of p63 ablation during embryonic epidermal development [35]. Despite the above studies, the relationship between DNp63a and p16 ink4a in cancer remains unclear. Consistent with the findings reported here, a recent study by Keyes et al. [2] demonstrated that overexpression of DNp63a bypasses oncogenic ras-induced senescence and drives tumorigeneisis in vivo. However, in contrast to our results, Keyes et al. reported that the level of p16 ink4a remains stable in oncogenic ras and DNp63a expressing cells, and concluded that the inhibition of senescence by DNp63a occurs via an additional pathway, but not through modulating p16 ink4a . This indicates the complexity of the pathways interacting with p63 family members, and underscores the need for additional studies to understand the role of p63 and its downstream effectors in tumorigenesis and senescence.
p63 is recognized as a regulator of epidermal cell fate and lineage commitment [42,43]. The pathways involved in regulating epidermal development as well as homeostasis and differentiation of postnatal keratinocytes continue to be targets of active investigation [6,28,44]. Much effort has focused on the identification of downstream pathways of DNp63a contributing to neoplasia, including NFkB/c-Rel, EGF-R and COX-2 [45][46][47]. Furthermore, the ability of DNp63a to inhibit transcriptional regulatory activity of other p53 family members may contribute to a loss of genomic stability that could enhance progression to malignancy [1]. Our laboratory has previously demonstrated that NFkB/c-Rel mediates the aberrant growth regulation observed in DNp63a-overexpressing keratinocytes [43]. Here, we establish that the ability of DNp63a to overcome senescence and drive malignancy is linked to its downregulation of the p16 ink4a /p19 arf pathways. As proof of principle that the loss of p16 ink4a /p19 arf can contribute to DNp63a-induced malignant conversion, an in vivo grafting study showed that, similar to lenti-DNp63a-expressing keratinocytes, p19 arf null keratinocytes formed malignant carcinomas in the presence of v-ras Ha (Figure 7). However, a significantly larger tumor volume was observed in tumors derived from lenti-DNp63a/vras Ha -expressing p19 arf null keratinocytes when compared to tumors derived from p19 arf null cells expressing v-ras Ha alone ( Figure 7B). Our results suggest that additional mechanisms, e.g. p16 ink4a or c-Rel, contribute to DNp63a -induced malignant tumor formation [45]. Future studies are required to dissect the relative roles of each individual downstream pathway influenced by p63 and their cooperative contribution to discrete steps of cancer pathogenesis.

Ethics Statement
All animal work was performed in accordance with NIH (National Institutes of Health) established guidelines, in accordance with accepted standards of humane animal care under protocols approved by the Animal Care and Use Committee of the Center for Biologics Evaluation and Research of the Food and Drug Administration.

Cell Culture
Primary keratinocytes and dermal fibroblasts were isolated from the skin of 1-2 day-old C57B1/6NCr mice, cultured and plated as described [20]. Keratinocytes were maintained at a final concentration of 0.05 mM Ca 2+ . Fibroblasts were cultured for 9 days in DMEM prior to use in grafting studies.
Adenovirus encoding DNp63a has been previously described [20]. The replication-defective Psi-2 retroviral vector encoding vras Ha was used to introduce a v-ras Ha oncogene at day-2 postplating as previously described [36]. The infection of adenovirus encoding Lac-Z or DNp63a was performed as previously described [20].

Lentivirus construction and infection
Lentivirus encoding human DNp63a was generated by ligating the complete coding region of DNp63a cDNA excised from pBMNiGFP-human DNp63a plasmid into lentiviral expression vector 960-X5-685 (pSICO-FerH-eGFP). The constructed pSICO-hFerH-human DNp63a was verified via restriction enzyme mapping and sequencing analysis. A similarly constructed lentiviral vector encoding GFP was used as a matched control. Lentiviruses were produced in HEK293T cells (Open Biosystems, Huntsville, AL). The producer cells were washed and incubated with standard keratinocyte medium 24 hours before harvesting (SAIC, Inc., Frederick, MD).
Keratinocytes, 3 days post-plating, were incubated for 3 hours in lentivirus-containing supernatant and 4 mg/mL polybrene (500 ml/60 mm 2 dish). Fresh medium was added at the end of the incubation. The induction of lenti-DNp63a was evaluated by western blot and immunofluorescent staining.

In vivo grafting and tumor sample collection
Keratinocyte cultures were infected with retrovirus encoding vras Ha one day prior to infection with lentivirus encoding either DNp63a or GFP. Cells were harvested six days following lentivirus infection and 4610 6 keratinocytes were mixed with 8610 6 cultured dermal cells and transferred to a grafting chamber freshly implanted onto the dorsum of a nude mouse [36,48]. The chamber was removed 1 week following grafting. Tumor volume was measured every 3 days following the initial appearance of a tumor. Mice were injected with 250 ml of 50 mM 5-bromo-29deoxyuridine (BrdU, Sigma Chemical, St. Louis, MO) 1 hour prior to euthanasia and tumor harvest to allow immunohistochemical analysis of DNA synthesis in the grafted cells. Tumor tissues were collected 4 weeks after removing the chambers and fixed in formalin-buffered saline.

BrdU incorporation
BrdU incorporation in the grafted tissue sample was detected by immunohistochemical staining using an antibody against BrdU (B44, BD Biosciences, San Jose, CA) and tissues were counterstained with Contrast Green (KPL, Inc., Gaithersburg, MD). At least 100 basal cells were scored for each sample; in the case of undifferentiated carcinomas 100 cells/field were counted. BrdU incorporation in vitro was detected by Fluorescent Activated Cell Sorting (FACS) analysis as previously described [32].

Senescence assays
For HP1c detection, culture dishes were fixed with methanol at 220uC and incubated with antibody directed to HP1c (MAB3450, Chemicon International, Billerica, MA), and p63 (4A4, Santa Cruz Biotechnology, Santa Cruz, CA). Cells were then incubated with a FITC-or rhodamine-labeled secondary antibody (Invitrogen) and viewed under a fluorescence microscope. Senescence-associated b-galactosidase (SA-b-gal) staining was performed as previously described [36]. Keratinocyte growth curve Keratinocytes were infected with retrovirus encoding v-ras Ha , followed by lenti-DNp63a or lenti-GFP. Cells were collected by trypsinization and counted using a hemocytometer (Hausser Scientific, Horsham, PA) at timepoints noted. At 14 days postinfection, the cultures were trypsinized and reseeded at 0.5610 5 cells per well, then counted as above at timepoints noted.

Immunoblot Analysis
Immunoblot analysis was performed using standard procedures. The following primary antibodies were used: p63 (4A4, Santa

RNA isolation and Reverse Transcription-PCR
Total RNA was harvested using TRIzol reagent (Invitrogen) and reverse transcribed (1 mg) using the SuperScript III first strand Figure 7. v-ras Ha -expressing p19 arf null keratinocytes display a tumor phenotype similar to v-ras Ha /DNp63a overexpressing keratinocytes. A, p19 arf null primary keratinocytes were infected with lenti-GFP or lenti-DNp63a in combination with v-ras Ha , and grafted onto the dorsal side of nude mice as previously described [36]. Wild type primary keratinocytes expressing lenti-GFP or lenti-DNp63a in combination with v-ras Ha were used as negative and positive controls. The final tumor phenotype was assessed at 5 weeks after grafting. B, Final tumor volumes at 5 weeks are presented as the mean tumor volume 6 S.E. * indicates a statistically significant difference between p19 arf null/v-ras Ha and p19 arf null/lenti-DNp63a/v-ras Ha groups at p,0.05. doi:10.1371/journal.pone.0021877.g007 cDNA synthesis kit (Invitrogen) with an oligo(dT) primer. Expression of p16 and p19 mRNA was determined by PCR amplification using primers specific for the mouse p16/p19 gene (p16: sense 59-GCTGCAGACAGACTGGCCA-39; antisense 59-GTCCTCGCAGTTCGAATCTG-39) [49]. p19: sense 59-GTCGCAGGTTCTTGGTCACT-39; antisense 59-ATGTT-CACGAAAGCCAGAGC-39) [50]. PCR amplification of GAPDH was used as a loading control. The PCR reaction program was as follows: denaturation, 94uC, 30 s; annealing, 53uC, 30 s; elongation, 72uC, 30 s for 35 cycles.